Trained Identification Model for On-Demand POI Ranking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online to offline services, such as on-demand transportation systems, face inefficiencies in determining relevant points of interest (POIs) based on predetermined manual rules, making it desirable to automate this process for improved efficiency.
Innovation Solution
A system and method that utilize a trained identification model, configured with historical transportation trip records, to determine and rank candidate POIs based on correlation probabilities, allowing for automatic and efficient recommendation of target POIs to users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If predetermined manual rules are used to determine correlative POIs, then the system implementation is simple, but the efficiency of determining POIs is low
Solution Approach 1:
The system uses historical transportation trip records to automatically train the identification model, enabling the system to self-improve without manual intervention. The model automatically learns from user selection patterns and refines POI recommendations over time.
Solution Approach 2:
The identification model is trained in advance using historical transportation trip records before being deployed for actual POI determination. This preliminary training phase allows the model to learn from past user behaviors and make accurate predictions when processing new address queries.
2Measurement precision
If manual rules are adjusted frequently to improve POI accuracy, then the POI determination accuracy improves, but the maintenance cost and time increase
Solution Approach 1:
The system incorporates user selection feedback from historical trip records to continuously improve the identification model. User selections serve as ground truth data that trains the model to better predict which POIs users actually want, creating a self-improving loop without manual rule adjustments.
Solution Approach 2:
The patent replaces manual rule-based systems with an automated machine learning model. The mechanical process of manually creating and adjusting rules is substituted with an automated statistical learning system that processes historical data and generates predictions automatically.
3Productivity
If a trained identification model is used to determine POIs, then the efficiency and accuracy of POI determination improve, but the device complexity increases
Solution Approach 1:
The identification model serves multiple functions: it determines POI recommendations, ranks them by relevance, and adapts to different user preferences. A single unified model handles various query types and user behaviors, reducing the need for multiple specialized systems.
Solution Approach 2:
The system uses historical transportation trip records as training data, creating a copy of past user behaviors and preferences. This historical data copy serves as the foundation for training the model, allowing it to learn from replicated patterns without requiring real-time human input.
Data Source
AI summary
The present disclosure relates to systems and methods for determining target search results associated with a target query. The method may include obtaining a transportation service request including a target address query from a user terminal, and determining a plurality of candidate points of interest (POIs) associated with the target address query. The method may also include identifying one or more target POIs based on the candidate POIs by using a trained identification model. The trained identification model may be configured to provide a correlation probability for each of the one or more target POIs with the target address query. The method may further include ranking some or all of the one or more target POIs to produce a ranking result based on the correlation probabilities, and transmitting the ranking result to the user terminal.